Channel economics after fees and refunds

The first number most channel reports show is cost per enquiry: what the channel cost divided by the people who got in touch. It is easy to collect and easy to compare, and that is why it misleads.

In April 2026 Klasna's lesson marketplace cost 20,000 UAH in the marketing line and brought 320 enquiries: 20,000 ÷ 320 = 62.50 UAH per enquiry. Paid social cost 150,000 for 400 enquiries, 375 each. On this number the marketplace wins six times over, and Uliana says "case closed".

An enquiry is only the first step of the path. Between it and a family that pays for months lie the first package, a possible refund, a second package and every cost the channel brings along the way. A channel can be cheap at the enquiry and expensive at every step after it.

So treat cost per enquiry as a starting point, never as the verdict. The question Hanna asks is the right one: what did each channel return after everything it cost?

The next number on most dashboards is spend per first package: the channel's marketing spend divided by the families who paid for a first package.

April 2026, counting only the marketing line:

• search: 90,000 ÷ 34 = 2,647.06 UAH;
• paid social: 150,000 ÷ 50 = 3,000;
• school partnerships: 60,000 ÷ 26 = 2,307.69;
• lesson marketplace: 20,000 ÷ 40 = 500.

This is better than cost per enquiry, because it counts people who paid rather than people who asked. And on this view the marketplace still wins by far.

But the number has two blind spots. It counts only what sits in the marketing line, so any cost booked elsewhere is missing. And it stops at the first payment, so it can't see refunds or whether the family ever buys again. For a business where the second package carries most of the contribution, stopping at the first package is stopping too early.

A channel's cost is everything it takes, wherever the books put it. The marketplace costs 20,000 UAH in the marketing line, but it also takes a 20% commission on every package booked through it, refunded ones included, and finance books that commission outside marketing.

Add it for the 40 first packages at 5,200 each: 40 × 5,200 × 20% = 41,600 UAH. The marketplace's cost so far is 20,000 + 41,600 = 61,600, three times what the marketing line shows.

Then count only families who kept their first package: 40 − 14 refunded = 26. Cost per kept family: 61,600 ÷ 26 = 2,369.23 UAH. Still a hair cheaper than school partnerships (60,000 ÷ 25 kept = 2,400).

So the marketplace's lead has shrunk from six times to almost nothing, and the most important part, second packages, isn't counted yet. The lesson: before comparing channels, list every cost each one brings, not just the line you manage.

A first package that is refunded leaves no contribution, and on a commission channel it can still cost money: the marketplace takes its 20% on refunded packages too.

April's refund rates differ sharply by channel:

• lesson marketplace: 14 of 40 = 35.00%;
• paid social: 6 of 50 = 12%;
• search: 3 of 34 = 8.82%;
• school partnerships: 1 of 26 = 3.85%.

A high refund rate usually means the channel brings people whose expectations don't match the offer. On a marketplace, families often book quickly, comparing many tutors on price, and leave if the first lessons don't feel right. A family sent by a school or a teacher arrives already knowing what Klasna is.

That is why counting first packages alone flatters some channels. Always subtract refunds before comparing, and look at the refund rate itself as a signal of fit: a channel that brings many first packages and many refunds is busy, not good.

A family that buys a second package has decided the lessons work. For Klasna that is where most of the contribution sits: a second package (16 lessons for 9,920 UAH) leaves 2,952, against 1,710 for a first one. So the sharpest channel number is full cost per family still paying after 90 days.

Second packages in April's 90-day window: partnerships 15 of 26, search 16 of 34, paid social 22 of 50, marketplace 5 of 40. And those five marketplace second packages went through the marketplace too, so commission was paid again: 5 × 9,920 × 20% = 9,920.

Marketplace full cost: 20,000 + 41,600 + 9,920 = 71,520. Per family still paying: 71,520 ÷ 5 = 14,304 UAH. Compare school partnerships 60,000 ÷ 15 = 4,000, search 90,000 ÷ 16 = 5,625 and paid social 150,000 ÷ 22 = 6,818.18.

The cheapest enquiry has become the dearest paying family. Nothing changed in the data; only how far along the path you followed it.

Hanna's question needs one number per channel: net contribution in the window. Contribution = kept first packages × 1,710 + second packages × 2,952. Subtract the spend and any commission.

• School partnerships: 25 × 1,710 + 15 × 2,952 = 87,030 − 60,000 = +27,030.
• Search: 31 × 1,710 + 16 × 2,952 = 100,242 − 90,000 = +10,242.
• Paid social: 44 × 1,710 + 22 × 2,952 = 140,184 − 150,000 = −9,816.
• Marketplace: 26 × 1,710 + 5 × 2,952 = 59,220 − 71,520 = −12,300.

The ranking is partnerships, search, paid social, marketplace. By cost per enquiry it was marketplace, paid social, search, partnerships: almost exactly reversed.

This is the number to put in a channel row of a plan, because it combines everything the earlier numbers left out: commission outside marketing, refunds, and the second package where the value sits. It still has limits, which the last card covers.

The last column of the April table says the most about the future: contacts in Klasna's CRM. For paid social, search and school partnerships it says yes. For the marketplace it says no.

That means a marketplace family's next package is booked on the marketplace again, with commission, and Klasna can't invite the family back itself: no email, no message about September, no offer when the child's next exam approaches. The families from Klasna's own channels can be reached directly.

This matters beyond the 90-day window. About 30% of spring families come back in September. Paid social's families come back to Klasna, so their return will improve social's numbers later. The marketplace's families come back, if at all, through the marketplace.

When you judge a rented channel, ask not just "what did it cost?" but "who owns the relationship afterwards?". A channel that keeps your customers keeps part of every future sale, too.

The April data supports some conclusions and not others. Supported: cost per enquiry ranks the channels almost in reverse of their 90-day net; the marketplace is the dearest per family still paying; part of its cost (51,520 of commission) never appears in the marketing budget.

Not supported: "stop paid social at once". Its net is negative, but 90 days is a window, not a lifetime: about 30% of spring families return in September, and social's families return to Klasna. Nor does anything say partnerships would bring the same result at five times the spend; what scaling each channel costs is a topic of the course Finance for Product People.

And mind the sizes: 26 to 50 families per channel in one month is thin. Trust the order; treat exact numbers as a first reading to check next spring.

How it fits together: follow each channel from enquiry past refunds to second packages, count every cost, and compare net contribution, then read it within its window. The next lesson, a case, does the same for Dana's one-person practice, where the budget is her hours.